Cyclophosphamide Metabolism in Children with Non-Hodgkin’s Lymphoma

Present treatment schemes for advanced-stage B-cell non-Hodgkin’s lymphoma (B-NHL) achieve cure in over 75% of children (1, 2, 3). In addition to the universal introduction of central nervous system prophylaxis, improved patient survival has also followed the use of fractionated cyclophosphamide in combination with cytarabine and high-dose methotrexate (4, 5, 6). Chemotherapy has been less successful in curing patients with central nervous system disease and bone marrow involvement. Although characteristic molecular rearrangements involving the juxtaposition of the MYC gene on chromosome 8q24 to one of the immunoglobulin-receptor-subunit genes on chromosome 2, 14, or 22 are widely recognized, the prognostic significance of these translocations remains unclear (7). B-NHL is a rapidly growing tumor, and recurrence occurring beyond 18 months after the start of therapy is exceptional (1, 5, 6, 8).

Cyclophosphamide remains one of the most widely used cytotoxic drugs (9) and is currently used in the treatment of many adult and pediatric cancers. It is a prodrug requiring metabolic transformation to generate active alkylating species (9, 10). The initial activation is mediated by hepatic cytochrome P-450 enzymes and is the major pathway of elimination of the parent drug (11). Hydroxylation at the carbon-4 position of the oxazaphosphorine ring produces 4-hydroxycyclophosphamide in equilibrium with the tautomer aldophosphamide, which spontaneously degrades to phosphoramide mustard and acrolein. Phosphoramide mustard is thought to be the active alkylating species (12). Alternatively, aldophosphamide may be oxidized to inactive carboxyphosphamide by aldehyde dehydrogenase (13). The other principal inactive metabolite, dechloroethylcyclophosphamide, is produced by a separate oxidative N-dealkylation reaction, which is also catalyzed by cytochrome P-450s (9, 11).

Previous studies have demonstrated a high degree of interpatient variation in drug metabolism, the clinical significance of which remains unclear (8, 9, 14, 15). To define the importance of this variation, we undertook an analysis of cyclophosphamide metabolism in 36 children receiving a uniform chemotherapy regimen for the treatment of B-NHL.

The pharmacokinetics and metabolism of cyclophosphamide were studied in 36 (15 female) children, median 8 years of age (range, 2–16 years) undergoing chemotherapy for B-NHL in the Children’s Cancer Unit, Newcastle upon Tyne, and the Department of Hematology, Royal Hospital for Sick Children, Glasgow. All of the patients had newly diagnosed B-NHL (Revised European American Lymphoma classification II; Burkitt’s lymphoma or high-grade B cell lymphoma, Burkitt’s-like). The diagnosis was confirmed histologically and after immunophenotyping with a panel of antibodies to B-cell antigens. All of the patients underwent staging by physical examination; computed tomography scan of chest, abdomen, and pelvis; full blood count; a unilateral iliac crest bone marrow trephine biopsy; and cerebrospinal fluid (CSF) examination. No child had bone marrow infiltration of greater than 25% of nucleated cells, or evidence of lymphoblasts in the CSF, and, thus, all were classified as Murphy stage II-III (8). Twenty-five children presented with abdominal tumors (69%), 5 presented with an inguinal mass (14%), 1 child presented with multiple soft tissue tumors, and 1 with a primary bone B-NHL. Four patients (11%) presented with bulky nasopharyngeal disease (Murphy stage II; Table 1). Successful karyotypic analysis was achieved in 23 cases providing cytogenetic evidence of the typical B-lineage translocations t(8;14)(q24;q32) or t(2;8)(p11;q24) in 64% of tumors. All of the children were treated in accordance with either SFOP protocol LMB 89/United Kingdom Children’s Cancer Study Group NHL 9002 (group B), or from 1996 onwards, SFOP protocol LMB 96/Children’s Cancer Group protocol 5953/United Kingdom Children’s Cancer Study Group NHL 9600 arm B1 (Table 2).

Written consent from parents and, when appropriate, from the subjects themselves was obtained before participation in the study. The study was approved by the joint ethical committee of the Medical School of the University of Newcastle upon Tyne and the Royal Victoria Infirmary, Newcastle upon Tyne, and the ethical committee of Yorkhill National Health Service Trust, Glasgow.

Patients were treated in accordance with the chemotherapy scheme outlined in Table 2. Initial chemotherapy with cyclophosphamide, vincristine, and prednisolone (COP) was given to achieve a gradual reduction in tumor volume and allowed for the resolution of metabolic or postsurgical complications before proceeding to more intensive therapy with COPADAM1. With the exception of COPADAM1, which was administered 8 days after starting chemotherapy with COP, the time interval between individual courses was 21 days, or until neutrophil and platelet counts had recovered to 1 × 10⁹/liter and 100 × 10⁹/liter, respectively. No patient received treatment with hematopoietic colony-stimulating factors.

Plasma samples were collected after the initial dose of cyclophosphamide on day 2 of COPADAM2. All of the subjects received the drug at a dose of 1 g/m² as a constant rate infusion over 1 h. Blood samples were obtained from an indwelling central venous catheter immediately before, and at 0.5, 1, 2, 4, 6, 12, 18, and 24 h after the start of the infusion. Concurrent medication varied in accordance with clinical needs and was not controlled. After drug administration, the concentration of cyclophosphamide and its principal metabolites were measured using a high-performance thin-layer chromatography photographic-densitometry technique (the lower limit of detection for carboxyphosphamide, dechloroethylcyclophosphamide, and ketocyclophosphamide was 3 μm; the lower limit of detection for cyclophosphamide was 1 μm; Ref. 9). Patients were deemed to have detectable metabolite if the concentration exceeded the lower limit of detection in one or more samples. Cyclophosphamide pharmacokinetics was assumed to follow a single compartment open model with first order elimination. Although a two-compartment model has been reported previously for cyclophosphamide, a one-compartment model has been found to be adequate in pediatric studies (10). Estimates of pharmacokinetic parameters were obtained by a maximum-likelihood method using ADAPT II (16). Area under the plasma concentration-time curves (AUC) for metabolites were calculated using the linear trapezoidal rule. Clearance was calculated as dose/AUC.

Patients were not preselected for this study and were enrolled sequentially as part of a 3-year investigation into intrasubject variation in oxazaphosphorine metabolism. All of the eligible patients were studied. This project was closed, and the clinical outcome of patients was analyzed according to their underlying diagnosis and therapy. No interim analyses were performed. The survival of this group of patients with NHL is consistent with that predicted from published reports (3, 7). Correlations between recurrence of disease and pharmacokinetic parameters were made using the Cox proportional-hazards model. The log-rank test was used to establish the significance of the detectable presence of carboxyphosphamide and dechloroethylcyclophosphamide.

Systemic clearance varied more than 2-fold with a median value of 3.6 liter/h/m² (range, 2.1–5.4 liter/h/m²), volume of distribution varied almost 3-fold with a median value of 0.56 liter/kg (range, 0.33–0.9 liter/kg), and half-life varied more than 5-fold with a median value of 2.7 h (range, 1.4–7.9 h). Carboxyphosphamide and dechloroethylcyclophosphamide were detected in eight children (22%). Measured metabolite AUC varied from not detectable to 206 μm·h and not detectable to 181 μm·h, respectively. Both carboxyphosphamide and dechloroethylcyclophosphamide appeared in all of the patients in whom metabolites were found. Ketocyclophosphamide was not detected in any patient sample.

After COPADAM2, only one patient did not require admission for the treatment of neutropenic fever. There were no toxic deaths. All of the children received red cell transfusions. The median duration of leucopenia (WCC, <1 × 10⁹/liter) was 12 days (range, 6–17 days), the median duration of neutropenia (neutrophils, <0.5 × 10⁹/liter) was 10 days (range, 7–15 days), and the median duration of thrombocytopenia (platelets, <50 × 10¹²/liter) was 8 days (range, 6–11 days). Subsequent chemotherapy was delayed beyond 21 days because of persistent myelosuppression in 14 patients (39%). Hematological toxicity was not significantly correlated with measured cyclophosphamide pharmacokinetic parameters or the production of inactive metabolites.

Two children did not complete the entire chemotherapy regimen as planned. In one case, treatment was discontinued after four courses of chemotherapy because of persistent hepatic dysfunction, and in another, the final three doses of methotrexate were withdrawn in the face of unexplained neuropathy. Both patients remain disease free. Pretreatment lactate dehydrogenase was documented in only 24 (67%) patients and was not subjected to further analysis.

All of the patients achieved a complete response after four courses of chemotherapy (confirmed at laparotomy in three cases). At a median follow-up of 43 months (range, 17–98 months) six children (four with abdominal and one each with inguinal and nasopharyngeal disease) have developed recurrent disease, giving an overall disease-free survival of 83%. Recurrences were at the site of the original disease in three patients, bone marrow in one, and central nervous system in two.

Univariate analysis demonstrated that tumor recurrence was inversely related to the clearance of cyclophosphamide (Table 3), which was less than 2.5 liter/h/m² in all of the treatment failures (Table 4). This relationship persisted after allowing for the effect of age, sex, and site of disease as possible confounding variables in a multivariate model (P = 0.008). Cyclophosphamide half-life was also inversely correlated with disease recurrence using the same multivariate model (P = 0.03). B-NHL recurrence was more common in children in whom carboxyphosphamide and dechloroethylcyclophosphamide were detected in plasma (P = 0.01 in each case; Fig. 1). Only one patient without detectable levels of circulating inactive metabolites developed recurrent disease.

Previous studies have defined the relationship between the systemic exposure of cytotoxic drugs, when used as single agents in children, and the clinical results of therapy (17, 18). Pharmacologically individualized treatment has also been shown to improve the outcome of children with B-cell lineage acute lymphoblastic leukemia when compared with dosing based on body surface area (19).

The data presented here suggest that variation in cyclophosphamide metabolism may influence the risk of B-NHL recurrence in children. This result is in agreement with a previous study that demonstrated an inverse correlation between the area under the concentration-time curve for cyclophosphamide and the duration of both disease-free survival and the incidence of cardiac toxicity in adults with breast cancer (20). In contrast, cyclophosphamide clearance was reported to be unrelated to the duration of disease-free survival in adults with Hodgkin’s disease, non-Hodgkin’s lymphoma, and breast cancer treated with multiagent high-dose chemotherapy followed by autologous bone marrow rescue (21). This discrepancy may have resulted from the use of differing multi-agent conditioning regimens and the inclusion of patients treated with a variety of previous therapies. Nevertheless, because cyclophosphamide is a prodrug, it might be predicted that patients with a low clearance would have less activation to the active metabolite.

A correlation between the production of the inactive metabolites carboxyphosphamide and dechloroethylcyclophosphamide and recurrence of disease was also observed. Although apparently self-evident, such a relationship has not been described previously, and a similar relationship could not be identified in a comparable study of ifosfamide (a structural isomer of cyclophosphamide that undergoes qualitatively similar metabolism) in breast cancer (22). This result suggests that the balance between the production of active and inactive metabolites by cytochrome P-450s is important in determining the treatment outcome. The absence of any significant correlation between the pharmacokinetics and metabolism of cyclophosphamide and subsequent hematological toxicity may have been obscured by the almost universal occurrence of severe myelosuppression in patients treated with this combination chemotherapy regimen.

The observation that cyclophosphamide clearance influences treatment outcome emphasizes the importance of potential drug interactions. Previous studies have shown that a range of drugs may alter cyclophosphamide metabolism in vivo. Cytochrome P-450-inducing agents such as dexamethasone and anticonvulsants result in an increase in clearance (23, 24). Conversely, concurrent treatment with fluconazole and thiotepa inhibit cyclophosphamide metabolism (15, 25). It is possible that the importance of these interactions in influencing the outcome of chemotherapy may have been under-recognized. In the present study, no administration of drugs known to interact with cyclophosphamide metabolism was documented.

Although cyclophosphamide clearance is increased after pretreatment with cytochrome P-450 inducers, it is unclear whether such treatment selectively increases the production of active alkylating species or simultaneously up-regulates inactivation pathways (23, 24, 26). Autoinduction of cyclophosphamide metabolism during repeated dosing is well recognized (14, 24), and modified schedules of cyclophosphamide administration may result in a greater degree of formation of active drug.

The findings of this investigation need to be corroborated in a larger prospective study, including a multivariate analysis to compare the influence of pharmacological variation relative to other clinical covariates of patient outcome. Nevertheless, these results suggest that pediatric patients with low rates of cyclophosphamide clearance, and those who produce significant quantities of inactive metabolites, are at greater risk of recurrence after current chemotherapy regimens for B-NHL. If such patients are identified by the monitoring of cyclophosphamide pharmacokinetics and metabolism, either an increased dose of cyclophosphamide or additional therapy with other agents should be considered. Alternatively, phenotyping or genotyping for the cytochrome P-450 enzymes involved in cyclophosphamide metabolism may identify, before treatment, patients in whom cyclophosphamide activation may be inadequate. This finding could have implications for the use of cyclophosphamide in other disease types.

We are grateful to Professor C. R. Pinkerton for his comments on the manuscript. We acknowledge the use of the North of England Children’s Cancer Registry without which the study would not have been possible.